Archive of March 2008

Holes in the Wall

By this I mean that I am able to run cables between my BT drop point all the way to the shed on the otherside of my house without having to drill through 3 internal and one external wall.

My problem however is how do I lay down large amounts of ethernet cable and still keep it looking tidy?

Putting blank covered socket boxes is one option but I would need at least three for this and a large amount of them, or any at all doesn't make for such a sleek install.

My other dilema is the mix between mains power and ethernet. In at least one position this would be very much required as the hole I am thinking about is a double socket extender box device with a switch hooked on the same ring for an outside light. I seem to be able suggest myself some shielding here.

This post is almost directly related to the one I did yesterday about the tests I did on my line and the position of my server in my shed from a couple of weeks ago. I am now off to search for the best way to go about doing this as it is most probably a job for this weekend. March 31 @ 06:00 PM | 0 Comments

Tweaking Be and the Be Box

Today I endevered to find out why I seem to get a rather low grade connection compared to the rest of the Be users. I had looked at other people's stats' for a while but did not consider actually looking into it and doing something about it.

Before I go on I am going to explain what I am using in terms of hardware/software. Firstly I am using a Speedtouch 585v6 (labelled as a BeBox) and DTM v7. DTM is a tweaking tool for various ADSL routers and can be found on this German website.

My current setup involves this router, a rather long extension lead (10/20meters), two microfilters and a badly wired master socket (thanks go to BT for that..). I have my networking equipment all around the same area on my desk, as its cheaperto run ethernet cables from here.

To start with I logged onto the router's web admin panel and pulled up the detailed connection stats. From here I noted down the listed speeds and line attenuation. This allowed me to set a benchmark before I did any work on my cables, or router. The benchmark gave me the following stats:

Bandwidth (Up/Down) [kbps/kbps]:    1,243 / 8,298
Line Attenuation (Up/Down) [dB]:    23.0 / 44.5

Firstly I removed the front panel from my NTE5 master socket and connected a different extension lead (this time a newer and lot shorter lead). This showed that my long extension lead was causing problems. Unfortunatly this time I forgot to note down the attenuation of the line, I did note down the speed though.

Bandwidth (Up/Down) [kbps/kbps]:    1,248 / 9,201

This showed a significant increase of 11%. Quite good I thought for a shorter cable. I decided that I needed to get a better speed however, but the limiting factor was the BT installed master socket. The contract states that your side of the network starts at the master socket and this is where all extensions must be plugged, or in the case of the NTE5 socket (which has a removable face plate) wired into. This basicly means that BT will charge you a rather large sum (I was lasted quoted £180 for an engineer call out).

From this point onwards (well until I get to the modem tweaking section), I am going to say that this is specific to my own premises and I do not take responsibility towards your own wiring. If you proceed to play around with and break your phoneline or damage your equipment I will not be responsible for it.

 So I proceeded to remove the whole of the master socket to examine and photograph the wiring (so if I forget how it was I won't damage my own equipment). I then went to the drop cable (in my case positioned in a box in my porch) and opened the box on the front of that. I have a splitter box which connects the outside wiring to my internal wiring.

From here I hooked up my master socket -sans 10 meters of bad quality cable running along my wall and connected it to the closest logical point. I then hooked up my modem and took stats again. This time it took the router a lot longer than normal to sync. From this I produced the following stats:

Bandwidth (Up/Down) [kbps/kbps]:    1,265 / 9,915
Line Attenuation (Up/Down) [dB]:    22.5 / 44.0

I think this is a pretty impressive given the actual length that is occupied at that end and also shows how quite terrible the cable is, and how pointless (in terms of ADSL runs) the distance does actually take.

Unfortunatly I done this at 4pm on a Sunday afternoon, so there was nothing logical I could get too to buy the parts required to adjust the connection this side.

That's it for that side of things. Next I decided to give DMT tool a go. DMT tool is an exellent peice of software which supports a wide range of routers through its many versions (although getting the latest version won't help you, the developer uses a versioning which you choose your router through each major version, then get the latest of that). In my case, I needed version 7.

I ran this on my Parallels virtual machine after setting the networking to Bridged on my wired ethernet connection. This is essentially a telnet front end which allows you to easily modify the ever so slightly hidden settings of these routers.

I am adjusting the SNR or Signal to Noise Ratio. The best place to look about this is on the Be and O2 users forum. As it is a forum I am going to assume it will be around for a long time.

Personally I have been slowly adjusting the bar in small increments and resyncing my router for the last couple of hours. I have been seeing a slow increase in the sync speed, although in rather small amounts each time. 

By the time this article is published I will have a new extension lead installed which is a lot shorter than the other one and more realistic length. From there onwards I plan to rerun through adjusting the SNR to see if I can get a stable 10mb/s connection (I currently have 9.53mb/s with the longer lead).

Future plans:

1. Rewire my master socket so it is adjacent to my drop line (the line that comes from the outside).

2. Reposition my router so that it is the closest it can be to the master socket.

3. Possibly upgrade the router, or setup a small embedded machine next to it which will pull the SMTP logs from it and tell me my network usage.

Evaluating Virtualisation Options

I have been slowly more and more disappointed with running VMWare on my server. For one I have never liked the idea of proprietary software for this kind of work and secondly I don't know enough Perl or C to make my own apps which fill this gap with what I currently have.

Most notably I only have the access (currently) of one server, so testing is almost out of the question unless I can do that testing in a couple of hours when that server would have to be offline.

This proves for quite a dilemma as I have to carefully work out solutions before I can consider rolling it out. I would not of course be considering this if the free VMWare Server fitted my needs exactly. Although it does to a reasonable degree I am not happy with the way the current beta has formed and having to use a significantly bodge setup causes me to not check in on my VM's as much as I should be able too. (I have to boot Windows in Parallels, then log onto the software which is slow and buggy)

So, my task for finding a better solution has come to finding something which I can install, understand and migrate current virtual machines over with less downtime than I am experiencing now. Plus, it would be nicer to have a more ergonomic software setup which is more responsive, and has less overhead.

Supposedly virtualisation is supposed to solve all the problems which I currently have. I have found the opposite, as with having less time on my hands and being overall more tired I don't sit around all day waiting for problems to happen. They happen and give me several hours of downtime and when I am told, I feel like an idiot for not noticing that there was a problem.

To outline my problem, I'd like:

• Unmodified guests. (I need to be able to use my current guests, most preferably)
• Open Source. (I feel locked down in VMWare)
• Remote Admin.
• Wide hardware support. (I mostly run ~2-3 year old hardware)
• Support for most x86 OSes.

The problem is that most CPU's (including the ones I own), are old and do not support newer technologies such as Intel VT-x. My MacBook does have this however and the effective speed of virtualised systems is very impressive, I can use my "Parallels" XP faster than I would if I were running it on a native box, and for the software which I use (which doesn't include games) this is perfect for me. My MacBook of course can't be used to virtualise servers though, as I carry it around with me.

The software I use has to support the current hardware I have access too and must be free (and have a wide variety of expansion modules).

Monitoring is my current problem. I am not told about downtime until something reasonably disastrous happens, such as the server being switched off for 12 hours, or running out of disk space, and I'm told by other people. This isn't so great in terms of reliability and the reputation of services run off my machines.

On the Remote Admin front I was looking for something that would provide a web panel and an actual app which I could run either on my main machine (my Mac), or on a virtualised GNU/Linux setup (which I could still run on my Mac, via X-Windows tunneled over ssh). I would obviously prefer to not be locked to a Windows client, as I am with VMWare and have reasonable web based access.

VMWare's commercial products are very good indeed and provide me with an working system, and if it were not for its current shortfalls I wouldn't be writing this now. That of course could be a shortfall within me, I could most probably put together an efficient monitoring system by just purely coding something using the VMWare API and some Perl. The problem is however that I don't know either of them at this current minute and I'd rather find something which I can setup instantly, rather than spend 3 months coding something which I doubt I will finish.

I have outlined the current virtualisation solutions below. I have tried to include everyone I could find which appeared to suit my purpose. Paid-for solutions are excluded.

(* I didn't bother looking into too much detail with QEMU. Expect a separate article on this amazing piece of work at a later date)

This of course isn't a definitive table. I've tried to focus on thinking in a server orientated manor. What it does show is that I could do with another reasonably identical machine to evaluate the whole thing and then roll out a dual system. One which uses more of a VPS setup and the other for running different OSes.

What I am trying to explain is that there appears to be three types of server-class virtualisation. One which requires VT/AMD-V for full performance and the other which uses the flexibility of GNU/Linux to modify the kernel and run it this way. The third and final uses the old-school way of doing things. Emulate many machines onto one with a reasonable performance loss.

This leaves a slight problem. Before I started this article, several hours ago in fact I planned to come out of the end with a solution to my problem which would not cost me a penny and would allow me to run many different types of operating system on a single box. I feel that I have ended up cutting this article far shorter than from what I had planned up in my head. I am however going to conclude this with a simple solution to other people's solutions...

For the virtual hosting of several machines, known as VPS'es, then OpenVZ (or Parallels commercial product based on the same kernel) is the best way to provide a base in which GNU/Linux machines can be resold.

For running many different types of OS on slightly older hardware then VMWare Server (version 1 for the time being until some of the problems of 2 are sorted) is most probably the best solution. It will allow you to run many, many types of virtual machine from a single box without requiring specialist hardware or particularly new hardware.

Xen. Seems to be the best solution if you have hardware support of Intel VT. Unfortunately, I can't test this.

Others, such as QEMU or some of Microsoft's products are a bit different and in both cases I haven't tested them or looked at them in detail.

For the mean time I am looking into rolling back to VMWare Server 1 on a lighter weight distribution (such as Debian) or CentOS rather than my current system which is an older install with Server 2 lumbered on top of it and setting up a monitoring system, such as Nagios.

Comments

On a trial basis I have decided to enable the comments function. I have noticed a sharp increase in visitors to this site and I think it may be time to give it a go.

I would be interested on your points of view on the subject, errors, ommissions, site tips, etc. Comments are enabled to be automatically aproved, so if you make a post it should go straight onto the site rather than me deciding to post it.

My plan is not to censor either, but I will, if nessisary.

Enjoy.

A book on blogging

I recently read a book on blogging.

It was a wrap but it did make me think of one thing. Don't post about pointless subjects. It covered this over a couple of pages, it mostly turned into a way of saying, look if you do this constantly it can be quite amusing, but its been done before, so don't dare do it.

Although it was a simple book, it was a rather nice read (for which it should have been about £9.99) whilst I was awaiting a lift in Guildford. People often tell me that book shops aren't library's. But isn't that just a silly proposition? I do actually buy books, but I never really read consistently. I now have many half finished books because unless I read it straight through, I get bored of it.